Prediction of aircraft fuel consumption based on machine learning algorithm

Abstract The shortcomings of traditional aircraft fuel consumption calculation methods include insufficient theoretical analysis, complex input parameters, and poor calculation accuracy. Through the analysis of aircraft motion equations and engine working characteristics, this paper screens out the key parameters affecting aircraft fuel consumption. Based on the real flight data, three machine learning methods, BP neural network, RBF neural network, and Support Vector Machine(SVM), are used for training, and a fuel consumption prediction model is established with key parameters as inputs. The prediction accuracy and characteristics of the three models are analyzed and compared. The aircraft fuel consumption prediction model achieves the operational requirements, and the accuracy in the flight profile reaches 5%. The trained model was applied to the planning of the transport aircraft mission profile, achieving a fuel consumption error of only 1.1% across the entire mission profile, resulting in a relatively accurate outcome.